Troubleshooting IPM Insights Issues

Use the information in this section to resolve issues related to IPM Insights.

Applies to

Enterprise Profitability and Cost Management, Financial Consolidation and Close, FreeForm, Planning (all application types), Sales Planning, Strategic Workforce Planning, and Tax Reporting.

Frequently Asked Questions

Review these frequently asked questions to see if they resolve your issue:

  • Does IPM Insights functionality require Hybrid mode?

    Yes, Hybrid mode must be enabled for IPM Insights functionality to work.

  • Can IPM Insights be set up on existing applications or is it available only for new applications?

    IPM Insights can be set up on existing applications.

  • Does IPM Insights functionality work with ASO and BSO cubes?

    Yes, IPM Insights functionality works with ASO and BSO cubes.

  • Is Auto Predict required for generating insights?

    No. You can also use predictions from external systems to generate prediction insights. With the Bring your own ML capability, you can train a Machine Learning model to predict the numbers, which can be used as the basis for prediction insights.

  • Do IPM Insights honor security?

    IPM Insights honor both member-level security and cell-level security, so planners see insights only for slices of data to which they have access. However, note that insights are not generated for a user when cell level security is defined on Year/Period dimensions.

  • Can IPM Insights be configured for parent members?

    Yes, you can select parent members in an insights job configuration. If the parent members are Dynamic Calc, there are additional steps required to generate insights. See Using Dynamic Calc Members in Insights and Predictive Planning, Auto Predict, and Advanced Predictions.

  • Can you configure insights only for value (such as $), or can you configure for units?

    You can configure insights for any measure or account—value or units.

Best Practices

Review these best practices to see if they resolve your issue:

  • Historical data: Prediction results are more accurate, the more historical data you have. There should be at least twice the amount of historical data as the number of prediction periods. For example to predict 12 months in the future you should have at least 24 months of past data. At the time of prediction if there is not enough historical data available, a warning is displayed. As a best practice, keep all historical data in an ASO cube and use that cube for analysis. Data quality is the most import aspect: good quality data can yield good insights and predictions.

  • Time granularity: When creating an insights job, create the lowest level of Period members possible so the greatest amount of historical data can be used.

  • Insight thresholds for different metrics: The recommended metric for Deviation is Mean Absolute Percentage Error (MAPE), which is a relative/percentage-based metric. Oracle recommends 10% as the suggested threshold value. However, this is only a recommendation. The most appropriate threshold value varies based on your historical forecast performance.

    For anomaly detection, the recommended methodology is Z-score method and the recommended threshold value is 3.

  • Historical forecast: For Forecast Variance and Bias insights, it is important that you have historical forecast numbers that are not adjusted for actuals. Once they are adjusted with actuals, you may not see a variance between the actuals and forecast.

  • Data storage for insights: Insights require large volumes of historical actual and forecast data. Consider storing this data in an ASO cube, especially for Anomaly and Forecast Bias insights.